Jharu – Free Open-Source Disk Cleaner for Developers

Jharu is a desktop disk cleanup application built in Rust (backend) and React/TypeScript (UI) using the Tauri framework, targeting developers and machine learning engineers on macOS and Windows. The core problem it solves is the accumulation of gigabytes of cache data from development tools—ML model downloads from Hugging Face, Ollama, PyTorch, and Whisper; package manager caches in npm, pip, conda, uv, yarn, pnpm, Cargo, Go, Maven, Gradle, and NuGet; Docker image and VM storage; JetBrains IDE caches; Xcode DerivedData; and leftovers from uninstalled applications. The app ships a per-platform knowledge base of 25+ known cache locations and classifies each item found, explaining what it is in plain language and rating its safety to delete. A distinctive feature is ML model intelligence: Jharu lists every downloaded model individually across Hugging Face, Ollama, Whisper, PyTorch, and Keras and distinguishes models that have actually been loaded from those downloaded once and never opened. This allows developers to reclaim unused models without risking breaking active workflows. For Python environments, Jharu sizes every virtualenv and conda environment and quantifies duplication across environments—the "torch is installed eleven times" problem. The app also analyzes the full disk as a treemap or list, classifying folders by type (system, applications, your files, app data, developer data), and can remove junk files (old installers, device backups, crash reports, mail attachment caches, large files untouched for six months). A daily capacity reading with least-squares forecasting projects when a drive will run out of space. The safety model is strict: every deletion moves data to the Trash or Recycle Bin (never bypassing it), every finding carries a safety rating, operating system locations are protected and cannot be cleaned, and each clean action is confirmed with a dialog stating the actual consequence. Registry "cleaners" on Windows and prefetch clearing are explicitly never offered, as the developer considers both ineffective. The app is free and open-source under the Apache 2.0 license with no telemetry, ads, or paid tier. Current version is 0.2.0, which shipped with per-model ML intelligence, Python environment analysis, a desktop cleaner, Windows uninstaller, deep clean treemap, and fill forecast. The main barrier to frictionless installation is the lack of code signing: on macOS users see "Apple cannot verify this app" and must right-click and confirm once; on Windows SmartScreen shows "Windows protected your PC." The developer is seeking sponsorships to cover Apple notarization ($99/year) and Windows code signing (~$100–400/year). Upcoming features include orphaned-data detection by comparing leftover folders against installed applications, duplicate-file detection, detailed Windows drive health (currently capacity only), and localized UI in Bengali and Hindi.

Jharu addresses a specific pain point for developers and machine learning engineers: disk space consumed by tools that leave large, opaque caches. Generic disk cleaners show folder sizes but do not understand what that data is; Jharu ships a curated knowledge base of 25+ cache locations across ML frameworks (Hugging Face, Ollama, PyTorch, Whisper, LM Studio, Keras), package managers (npm, pip, conda, uv, Cargo, Maven, Gradle, NuGet, Go), containers (Docker), IDEs (JetBrains), and build tools (Xcode). A distinctive feature is its ability to distinguish ML models that have been loaded from models downloaded and never used—allowing developers to reclaim storage without breaking active workflows. The app's safety model is conservative: every deletion moves to the system Trash (never permanent), every item carries a safety rating and staleness detection, and locations Jharu cannot read are reported rather than hidden. The developer positions Jharu as a direct alternative to CleanMyMac (paid, macOS-only, lacks ML model understanding) and CCleaner (freemium, no developer-specific intelligence). The current limitation is that Jharu lacks code signing on both platforms—users see security warnings on first launch—and the developer is explicitly seeking sponsorships to cover Apple notarization ($99/year) and Windows code signing costs (~$100–400/year). The roadmap includes orphaned-data detection, duplicate-file detection, and localization into Bengali and Hindi.

No comments yet. Be the first to share your thoughts!